That seems like a reasonable thing to want to have.

I don't think we know how to implement such a thing as an intrinsic. The instruction x86 would use will trap if the quotient is too big, regardless of whether you use the quotient or not.

In that case, I personally would very much want to keep math/bits.Div{32, 64} simple and fast (i.e. as it is now) and perhaps include Rem{32, 64} sometime in the future, if desired.

Ok, we can decide if we want Rem also (and how to do it, if we do) for 1.13.

Note that you can do Rem32 yourself using the built-in 64-bit divide.

For the 64-bit case, I not sure if you can get around the need to issue two division instructions to get the remainder when hi >= y. But an implementation you can use now for that case is:

func Rem64(hi, lo, y uint64) (rem uint64) {
  _, rem := bits.Div64(hi%y, lo, y)
  return rem
}

@bmkessler : Very nice.

Based on comments above, sounds like decision is keep panicking and leave Rem32/Rem64 for Go 1.13.

Assuming here that Div64 is new in this release, which I think it is.

Too late to add new API to 1.13.

Change https://golang.org/cl/197838 mentions this issue: math/bits: add Rem, Rem32, Rem64

Thank you!

Change https://golang.org/cl/217127 mentions this issue: doc/go1.14: mention new math/bits functions Rem, Rem32, Rem64